Balisk · Home

AI search optimisation · GEO

Diagnosis, implementation and maintenance.

We audit, fix and measure. We do the work and you see every change.

Balisk offers a single service in three phases: a technical diagnosis of about 6 weeks, a 90-day implementation and month-by-month maintenance. We audit the whole chain by which an AI comes to name a business: access, reading, indexes, queries, passages, sources, entity and answer. Then we apply the changes, and each one stays in your portal with its date and its proof.

01 · Diagnosis

We audit all eight links.

We review the whole path, from crawler access to the final answer, and tell you where your business drops out and why.

A magnifying glass scans the website, ticks every correct box and flags the failing one in red.

Technical scope

What we audit, link by link.

  1. Access

    What your server answers to each AI crawler, checked against your robots.txt, your CDN, your firewall and, if you give us access, the server logs.

    • OAI-SearchBot
    • ChatGPT-User
    • Claude-SearchBot
    • PerplexityBot
    • Googlebot
    • Bingbot
  2. Reading

    Whether your key content is in the initial HTML or depends on JavaScript, the response speed, and whether your structured data validates and matches the visible text.

    • initial HTML
    • JavaScript
    • JSON-LD
    • schema.org
  3. Indexes

    Whether your key pages are in Google, Bing and Brave, with a sitemap and truthful update dates.

    • Google
    • Bing
    • Brave
    • sitemap
    • update dates (lastmod)
  4. Queries

    Your customers’ real questions in several AI tools (a fixed panel of about 50), measured with their 95% margin (about 13 points either way), and the fan-out searches the AI runs from them.

    • fan-out searches
    • 95% margin
    • market language
  5. Passages

    Whether your pages answer directly, with facts, figures and dates the AI can cite.

    • direct answer
    • one page per intent
    • HTML tables
  6. Sources

    Which websites each AI cites when it talks about what you offer, and which ones name your competitors and not you.

    • press
    • reviews
    • directories
    • forums
  7. Entity

    What each AI believes about your business against reality, and the consistency of your data across profiles, directories and Wikidata.

    • name, address and phone
    • sameAs
    • Wikidata
  8. Answer

    In how many answers you appear, whether you are recommended or only cited, with what data and against whom.

    • presence
    • share of voice against competitors
    • recommendation or citation

You get 3 priorities, ranked by impact.

A report ends in three ranked cards; the first, red and in front, is the priority.

Deliverables

Six deliverables and a plan.

  • Executive report

    Your 3 priorities, ranked by impact.

  • Truth sheet

    What each AI says about you, what is correct and the source behind each error.

  • Visibility matrix

    Question × AI, with presence and margin, against your competitors.

  • Source map

    The websites that carry weight in your market, ranked by opportunity.

  • Technical issues

    Each one with its proof and the exact change.

  • 90-day plan

    Every action with who does it, when and how it is checked. We present it to you in a meeting.

Phase sheet

How long does it take?
About 6 weeks.
From the moment we validate the questions with you. Over the following two weeks we run the whole question panel, because a single AI answer tells you nothing.
When does it make sense?
When your customers ask before choosing.
When they compare options before deciding and can ask an AI about what you offer. The diagnosis tells you, with data, what each AI answers today, why it names others and which link fails first. It is the starting line: without it we could not measure later what has changed.
What do we need from you?
A meeting, your data and read-only access.
A first meeting, your correct business data and, if you can, read-only access to Search Console, Bing Webmaster Tools, Analytics and your server or CDN logs. Without that access the diagnosis still goes ahead, but access and indexes are checked in less depth.
What do you see in your portal?
The report and every measurement.
The report to download, every measurement with its margin and the list of priorities, which then becomes the implementation plan.

Frequent questions

How does it differ from the free checker?

The checker looks, in seconds and from the outside, at whether AI tools can read a page of your website. The diagnosis audits the whole chain: it asks the AI tools about what you offer, measures in how many answers you appear and whom they name instead, traces each error to its source and gives you the plan.

Can I start directly with the implementation?

No. The diagnosis is the starting line: without it we would not know what to change first, nor could we show you afterwards what has improved.

02 · Implementation · 90 days

We apply it in 90 days.

We fix your website, your business data and your external sources, starting with what has the most impact.

The pieces drop and fit into your website one by one; the last one, in red, is the key change.
  1. Days 1–30 Technical base Access, reading without JavaScript, indexes and entity data.
  2. Days 31–60 Content and data Pages that answer the searches AI runs.
  3. Days 61–90 External sources Profiles, directories, reviews and media, all consistent.

Technical scope

What we change, inside and outside.

  1. Access

    robots.txt and CDN and firewall rules that let verified crawlers through. The decision on training is yours, in writing.

    • robots.txt
    • CDN
    • WAF
    • Google-Extended
  2. Reading

    Key content in the initial HTML, with no dependence on JavaScript, and entity structured data with stable identifiers and every value visible on the page.

    • initial HTML
    • JSON-LD
    • stable @id
    • visible text
  3. Indexes

    A sitemap with truthful dates, IndexNow to notify Bing and other search engines of every change, and presence checked in Google, Bing and Brave.

    • sitemap
    • IndexNow
    • Google
    • Bing
    • Brave
  4. Content

    Literal changes page by page: the direct answer first, one page per intent, facts with a figure and a date, HTML tables and headings in your customers’ own words.

    • direct answer
    • dated facts
    • HTML tables
    • hreflang
  5. External sources

    We ask for the wrong data published about you to be corrected and give you a plan for press, reviews, directories and forums, with the texts ready and no deceptive tactics.

    • press
    • reviews
    • directories
    • forums
  6. Entity

    Your data identical on your website, on your Google, Bing and Apple business profiles, in directories and in Wikidata, where you meet their criteria.

    • business profiles
    • directories
    • Wikidata
    • sameAs

A typical audit, before and after

Links passed
Links passed: before, 1 of 8; after, 8 of 8.
Consistent sources
Consistent sources: before, 2 of 9; after, 9 of 9.
Queries covered
Queries covered: before, 3 of 14; after, 12 of 14.
Before: one broken link is enough for AI to leave your details out. After: all eight links checked, each with its proof. Reference counts from a typical audit, not a client’s data.

Your details, the same everywhere.

Your website, your map listing, a directory and your reviews lined up with the same red stripe.

Deliverables

Each phase, with its delivery.

  • Days 1–⁠30 · Technical base

    Access, reading without JavaScript, indexes, entity data and IndexNow, applied and verified.

  • Days 31–⁠60 · Content

    The literal list of changes per URL and the new pages by intent, in the languages of your website.

  • Days 61–⁠90 · Sources

    Your entity data identical everywhere and the plan beyond your website, with the texts ready.

  • Closing report

    What was done, what has changed, with what margin and what still cannot be attributed.

Phase sheet

How long does it take?
90 days, in three phases.
Three 30-day phases: first what stops you being read (access, reading and indexes), then content and data, and finally entity and external sources. We follow that order because improving a page the AI cannot read is pointless.
Who is it for?
For those who already have their diagnosis.
For businesses with a recent diagnosis who want us to apply the changes and verify them one by one, on their website and beyond.
What do we need from you?
Your approval and a way to apply changes.
Your written approval before we publish each change, and one of two routes: you give us access to your website and we apply it, or your web team applies our exact text and we check it.
What do you see in your portal?
Every change and its status.
Every change with its status (proposed, approved, applied and verified), its date, where it was made and proof that it is right.

Frequent questions

Do I have to give you access to my website?

It is not mandatory. If you prefer, we hand your web team the changes ready to apply and check that they come out right. What matters is that someone can apply them.

And if something goes wrong after a change?

Before each change we apply ourselves, we save a copy, and if something goes wrong we restore it as soon as we detect it. If your web team applies it, with each change we give them the exact steps to undo it.

Do you use hidden text or tricks to fool the AI?

No. Hidden text and instructions concealed for the AI go against Google’s spam policies, and fake reviews are banned in the EU. We work only with what can be shown: true, consistent, sourced data.

03 · Maintenance

We maintain it monthly.

Every month we check that nothing has broken, apply improvements and review your data. You see it all in your portal.

A twelve-month calendar with every month checked; the current month lights up in red.

Technical scope

What we monitor and improve.

  1. Technical monitoring

    Blocked crawlers, a robots.txt that changes, pages that stop being readable without JavaScript or broken structured data. The basics are checked every day and the rest every week. If something serious breaks, we tell you within one working day of detecting it.

    • robots.txt
    • initial HTML
    • JSON-LD
    • indexes
  2. Monthly drift

    With a monthly sample of questions, run by hand in ChatGPT, Google and Gemini: model changes with their date, sources that enter or leave the answers and new errors about your business.

    • model changes
    • sources
    • accuracy
  3. Verified improvement

    At least one verified improvement a month on your website, your profiles or your sources, always with your approval. If the approval comes late, that improvement moves to the following month.

    • approval
    • verification
    • backup before each change
  4. Quarterly panel

    The full question panel with its margin and a significance test against the previous quarter.

    • 95% margin
    • real change or noise
    • same questions

The maintenance rhythm

  1. Every week A full technical review, with a daily check of the basics. If something serious breaks, we tell you within one working day of detecting it.
  2. Every month The drift, from a hand-run sample of questions, at least one verified improvement (if your approval comes late, it moves to the following month) and a note with your Search Console, Bing and Analytics data.
  3. Every quarter The full panel with its margin, the significance test and a meeting to decide the focus.

Phase sheet

What do you receive?
A monthly note and verified improvements.
At least one verified improvement every month, always with your approval (if the approval is delayed, the improvement moves to the following month), and a monthly note with what was done, the drift and your own Search Console, Bing and Analytics data. Every quarter, the full measurement and a meeting to decide the next focus.
How long does it last?
Month by month.
Month by month, after a minimum of 3 months. To end it, 30 days’ notice is enough.
Who is it for?
For those who have already implemented.
For businesses that have completed the implementation and want to keep what they achieved. AI answers change with every model and every new source, and a diagnosis is a snapshot that expires.
What do you see in your portal?
The progress and every change.
The log of everything done, the alerts and how they were resolved, each month’s note and the evolution of your presence in AI answers, with its margin.

Frequent questions

Why not a monthly report of mentions?

Because from one month to the next the difference usually falls within the margin of error: we would be showing you noise. Every month you see what was done and verified, and the drift; every quarter, the full measurement with statistical rigour.

What happens if we stop working together?

We hand over your data in an open format, withdraw our access and leave you a note with what remains pending.